When a user uploads an image attachment in the chat interface and asks the agent about it, the agent replies that it hasn't received the image. The attachment never reaches the agent's context.
Steps to reproduce:
- Open a conversation in the production environment.
- Upload an image file as an attachment.
- Ask the agent a question about the image (e.g., "What's in this image?").
- Observe the agent's response — it states it hasn't received any image.
Environment:
- Production environment
- Models tested:
deepseek-v4-flash and gpt-5-luna
- Both models exhibited the same behavior
Expected behavior: The agent should receive the uploaded image and be able to reference it in responses.

When a user uploads an image attachment in the chat interface and asks the agent about it, the agent replies that it hasn't received the image. The attachment never reaches the agent's context.
Steps to reproduce:
Environment:
deepseek-v4-flashandgpt-5-lunaExpected behavior: The agent should receive the uploaded image and be able to reference it in responses.